Abstract
Bayesian analysis was adopted in combination with hierarchical (population) modelling to estimate population and individual insulin sensitivity SI and glucose effectiveness SG with full (30 sample) and two reduced (12 sample and 13 sample) sampling schemes. After overnight fast, 65 Caucasian subjects with newly presenting Type 2 diabetes according to WHO criteria underwent insulin modified IVGTT. The 13 sample scheme was preferred to the 12 sample scheme and gave accurate estimates of population SI, individual SI, but not population SG and individual SG. We conclude that the addition of a single sample with insulin modified IVGTT substantially improves accuracy of the calculations of insulin sensitivity.